audit-verification

To check for unverified task completion claims, compare claimed results against actual build, test, and typecheck run evidence to detect gaps. This non-blocking process emits structured signals for human review without halting progress in task-driven AI software pipelines.

The best way to detect unresolved references is to diff claimed changes against files the build would reject. This audit identifies mismatched file references and reports them as structured signals for later human review.

A non-blocking build verification hook works by using a deterministic Bash script to queue proposals and emit signals without halting execution. It runs on demand to surface unverified completion claims for later review.

Yes, you can run verification checks without blocking your release pipeline. The audit applies to task-driven AI workflows in software QA by emitting signals on demand via a zero-LLM Stop hook for later review.

No, you do not need an LLM to audit unverified completion claims. The verification uses a deterministic, non-LLM Bash hook to queue proposals and emit signals for later human review on the cheapest tier.

You should avoid using a non-blocking signal audit when your workflow requires immediate hard failures to prevent bad merges. This tool never blocks progress, only queuing proposals and emitting signals for on-demand human review.